Periodic scatterometry targets are used to obtain accurate measurements of target features. Such targets include massive arrays of uniformly constructed and uniformly spaced periodic features arranged to provide the best possible targeting information. For example, periodic gratings may be used as targets as may other periodically configured higher dimensional target arrays having uniformly spaced and sized metrology features.
Current scatterometry overlay (SCOL) targets are non-design-rule targets, which include features or spaces as large as 400 nm. A typical SCOL target consists of several cells, each consisting of two gratings (one in each of the layers between which the overlay needs to be measured). An example of a grating in one of these layers is seen in FIG. 1A. In this grating the typical size of a feature or a space is hundreds of nanometers (pitch 103), in contrast with design rule features, which are tens of nanometers in size. The features of a SCOL target are sometimes segmented for better process compatibility as seen in FIG. 1B. The fine pitch 103B of the segmentation can be as small as tens of nanometers, similarly to the design rule of the device. However, the spaces in such a segmented target are still of size of hundreds of nanometers (pitch 103A), and therefore this target may become distorted and noisy because of process effects. This may require spatial averaging of the target, which by itself limits the target size from below to be the spatial averaging size. Furthermore, it is well known that 1st order SCOL technologies tend to be sensitive to asymmetric grating imperfections, and that, in cases where one of the gratings reflects significantly more light than the other, the sensitivity to overlay is low. Finally, to gain more sensitivity to overlay, current SCOL technologies require the printing of more targets on the wafer (with varied programmed offsets). This increases the real-estate of the targets and the COO (cost of ownership) of the metrology tool.
Another aspect of current 1st order SCOL technologies is that they have TIS (tool induced shift) and TIS3s (tool induced shift 3-sigma—a variability value relating to the TIS) that result from non-zero illumination asymmetry. To reduce TIS and TIS3s one needs a variety of error-prone calibration techniques which lead to a residual TIS and TIS3s. Another disadvantage of current 1st order SCOL technologies is that there is no direct per-pupil-coordinate weight that is strongly correlated to accuracy.
Another aspect of current 1st order SCOL technologies is that they are based on comparing signals performed at different times (signals that correspond to pupil images of different target cells). These signals experience different system noise which needs to be removed. The sensitivity of the overlay to miss-handling the system noise is significant, and leads to very tight tolerances on this parameter.
Periodic targeting structures typically feature two layers of similarly oriented periodic gratings formed one over the other. Typically, the layers are designed with a specified predetermined offset with respect to each other. This enables scattering signals to be generated when illuminated by a light beam. A comparison of the actual signal produced with the expected scattering signal enables highly accurate overlay metrology measurements to be made. Optical metrology targets can also comprise of single gratings and/or gratings in a single layer, for example in optical metrology of critical dimension or in overlay optical metrology having targets positioned side by side.
Current SCOL target designs comprise of finite size cells 90 which include gratings 80, 85 of a defined pitch. The number of gratings and their position depends on the specific SCOL technology. For example, in 0th or 1st order SCOL, a target comprises of several cells, each cell comprising of two gratings in two different layers. In the single patterning case, for instance, the two layers are positioned on top of each other, with, possibly, several film layers in between. Relative offset 75 of the grating position includes a programmed offset (pof) and the overlay (ovl). The main SCOL paradigm is that the asymmetry in the cell is solely due to the total offset and so that rotating the target by 180° is equivalent to negating the sign of the total offset. This basic assumption leads to a variety of algorithms that take as input the asymmetry signals of various cells with different values for pof, and use it to extract the overlay.
FIGS. 1C-1H schematically illustrate prior art cells in standard scatterometry overly targets and their deficiencies. FIGS. 1C, 1E and 1G are top views, FIGS. 1D, 1F and 1G are cross sectional views. FIGS. 1C and 1D illustrate a target 90 having one cell with edges 70 and a grating 85 upon a layered target area 60. Generally, targets 90 are not symmetrical with respect to a 180° rotation 74 about a central axis 73 perpendicular to the target's face, due to production considerations. As illustrated in a depiction of one cell 90 in FIGS. 1E and 1F, a lower grating 80 is positioned in the bottom layer of a target area 60 and an upper grating 85 is positioned on the top layer of the layered target area 60. A perimeter 70 depicts the cell edges and an axis 73 that is perpendicular to cell 90 and central with respect to cell edges 70 is depicted too. However, prior art cells 90 do not exhibit symmetry for a 180° rotation 74 about central axis 73, mostly for reasons relating to the manufacturing of the targets. FIGS. 1E-1F illustrate targets having a total offset 75 that is introduced as the sum of difference of a programmed offset (pof) and the overlay (ovl). FIGS. 1G and 1H illustrate a zero offset 75 case. In the top view the pictorial representation shows only the upper grating since the lower grating is hidden by it (they have the same critical dimension—CD in this pictorial representation). FIG. 2F illustrates a high level schematic top illustration of a prior art cell with a two-dimensional target that is asymmetric with respect to 180° rotations 74A, 74B about axis 73 in both dimensions. Common to all these prior art targets is that the target cells are not symmetric with respect to cell edges 70 when subject to 180° rotations about a central perpendicular axis 73.
Scatterometry overlay (SCOL) technology, as illustrated e.g., in WIPO publication no. WO 2004076963, measures an overlay error between congruent targets in different layers by measuring the interferences of reflected diffraction orders from the targets.